Integrated-approach-based Automatic Building Extraction

This paper presents an approach to automatic building extraction from large-scale aerial images. It is based on the integration of multiple information, knowledge and appropriate methods, and hierarchically consists of three major parts: building detection, building segment extraction, 3D feature matching and building modelling. Buildings are detected by segmenting DSMs using grey-scale mathematical morphological operations, and by incorporating texture and shadow information for refinement and verification. The reconstruction process is focused on the detected building regions, and is accomplished by a bottom-up process: from primitive feature derivation, building segment extraction, segment matching, plane face constructing to 3D modelling. The approach uses a generic polyhedral model for extraction, so it can deal with buildings with different shapes and structures. Test results of the approach are presented in the paper.